INFRARED CAMERA MEASUREMENT CORRECTION FOR PULSED EXCITATION
WITH SUBFRAME DURATION
TECHNICAL FIELD
[0001] This invention generally relates to infrared camera measurement systems, and more particularly relates to infrared camera measurement systems used in thermography.
BACKGROUND OF THE INVENTION
[0002] Active thermography is used to nondestructively evaluate (NDE) samples in order to detect sub-surface defects. It is effective for uncovering internal bond discontinuities, delaminations, voids, inclusions and other structural defects that are not detectable by visual inspection of the sample. Generally, active thermography involves heating or cooling the sample to create a difference between the temperature of the sample and the ambient temperature and, then observing the infrared thermal signature that emanates from the sample as its temperature returns to the ambient temperature. Pulsed thermography is widely used in the nondestructive evaluation of component parts used in aerospace and the power generation industry.
[0003] An infrared (IR) camera is typically used for thermography because it is capable of detecting anomalies in the infrared spectrum irradiated from the sample. These anomalies are produced in the cooling behavior of the sample when sub-surface defects are present because the sub-surface defects block the diffusion of heat from the surface of the sample. In particular, sub-surface defects cause the surface immediately above the defect to cool at a different rate than that of the surrounding (defect-free) areas. As the sample
cools, the IR camera captures and records an infrared image of the sample, creating a sequential, time record of the sample's surface temperature.
[0004] In performing thermography, it is typically assumed that the integration time, (i.e. the time during which photons are collected by the focal plane array (FPA) detector of the infrared camera) , occurs simultaneously with the frame synchronization (hereinafter frame sync) or vertical sync signal. In fact, in many high performance IR cameras typically used in NDT applications, the integration time precedes the frame sync by a percentage of the frame sync period. For example, the integration for a given time frame may occur during the outputting of the previous frame (integrate while read mode) . The precise time at which the temperature measurement is made may differ from the apparent time (based on the frame number) by a significant amount. This difference is especially acute in the earliest post- flash frames .
[0005] The present invention uses an infrared camera to accurately measure the onset of the flash event pulse with respect to the frame sync signal. The present invention also uses the infrared camera to measure the duration of the flash pulse event. This is accomplished by detecting a slight disturbance in the pixel values in the frame that is read-out concurrently with the occurrence of the heating pulse.
BRIEF DESCRIPTION OF THE DRAWINGS
[0006] FIG. 1 is an idealized timing diagram of a pulsed thermography measurement;
[0007] FIG. 2 is a typical, non-idealized, timing diagram of a pulsed thermography measurement;
[0008] FIGS. 3A and 3B is an image formed by subtracting pixel values from Frame 0 from pixel values taken from a prior frame for a 1 ms flash pulse;
[0009] FIGS. 4A and 4B is an image formed by subtracting pixel values from Frame 0 from pixel values taken from a prior frame for a 4 ms flash pulse
DISCLOSURE OF THE INVENTION
[0010] Now referring to Figure 1, in most instances, the effect of the timing skew (error) between the flash pulse and the frame sync signal is negligible because, in active thermography NDE applications, temperature does not typically change significantly over an interval defined by two consecutive frames. However, in early post-flash frames of a pulsed thermographic NDT measurement, the error can be considerable. The significance of the error can be illustrated by considering equation (1) below which is the solution to the ID diffusion equation for surface temperature of a sample following application of an instantaneous, uniform heat pulse:
[0011] wherein T is the temperature of the sample after the application of the heat pulse,
[0012] To is the temperature prior to applying the heat pulse,
[0013] Q is the heat pulse energy per area of the sample,
[0014] e is the sample's material property of thermal effusivity, and
[0015] t is the time after application of the heat pulse.
[0016] Using an infrared camera necessitates that the temperature measurements acquired by the camera are acquired at discrete time intervals separated by the frame period τ which is I/frame rate. Typically, it is assumed that the heating pulse 19 occurs at the rising edge of the Frame Sync signal for frame number N=O (as depicted in Figure 1) and that each frames' FPA measurement 21 coincides with the rising edge of its respectively associated Frame Sync signal (the rising edge of each frame N occurs at time t=NT after the heating pulse) .
[0017] In this ideal situation, the surface temperature measurements in the ID case is:
T(N)-T0= & (N>0) (2) e\π^Nτ
[0018] This ideal pulse thermography measurement timing discussed above is set forth graphically in Figure 1.
[0019] Although the idealized assumptions referred to above are often sufficient for most NDT measurements, in certain NDT applications, they can lead to significant errors. For example, in most thermographic NDT measurements, the heating pulse 20 (see Figure 2) and the onset of the FPA integration periods 22, 24 and 26 are not coincident with the rising edges of the Frame Sync signals 28, 30 and 32. Figure
2 shows a significant (but typical) time skew between the heating pulse 20, FPA integration periods 22, 24, 26 and the rising edges 28, 30, 32 of the Frame Sync signals. If the flash heating pulse is offset from N=O Frame Sync by tfiash and the integration time for frame N=I is offset by time tint from N=O, then the correct times from heating pulse to measurement are t= (N-I) T+tfiaSh - tint. Accordingly, a corrected equation taking into account this offset is as follows:
[0021] The differences produced between equation (2) and equation (3) is small for large values of N. However, for early frames (frames close to N=O) the correction set forth in equation (3) has a significant impact in analyzing the thermographic data, such as when analyzing the data using log-log plots, TSR processing, or numeric time derivatives.
[0022] In practice, the integration period tint, 22, 24, 26 for a camera is usually synchronized to the Frame Sync and to the integration offset time tint. This information can typically be obtained from the IR camera manufacturer for the various modes of operation of the IR camera. However, the timing, tfiash , of the flash heat pulse 20 is not necessarily synchronized with the Frame Sync. Also, flash offset time, tfiash can vary as a function of any number of factors some of which include flash lamp system settings, variations in state of computer hardware and software, and changes in camera settings such as frame rate and frame size.
[0023] It is desirable to have a reliable method of measuring the flash pulse offset time, tfiash , and flash pulse duration 23.
[0024] An embodiment of the present invention is effective for detecting the onset of the heating flash pulse and flash pulse duration by monitoring the pixel values generated in the frame that is being read-out at the time that the heating pulse is generated. The pixel values read¬ out during a given frame, are usually comprised of image values that were captured during the integration period of the prior frame (see Figure 2 wherein N=I integration period 22 (which is output in Frame N=I) is conducted during the time that Frame N=O is read-out) .
[0025] Camera manufacturers do not design their cameras with the intention that the integrated pixel values will change between integration periods. However, the flash heating pulse used in pulsed thermography is intense and influences (i.e. increases) the pixel values that are read¬ out (simultaneously with a flash event) by a small amount. The increase is a function of the flash intensity. The magnitude of this pixel value disturbance is typically small (approximately 40 parts out of 10,000) .
[0026] The disturbance is most easily observed by subtracting the pixel value collected from one or more pre- flash frames with the corresponding pixel values collected from the frame being read-out during the heating pulse. The pixels in a frame are read-out sequentially in time (serially) . A pixel's value is only disturbed if a flash is occurring simultaneously during that pixel's read-out. For example, if a frame, N=O is read-out during a heating pulse (see Figure 2), then subtracting frame N-I (pixel-by-pixel)
from the corresponding pixels in frame N, creates an image that highlights the pixels disturbed by the heating pulse (see Figures 3A-4B) . The flash heating pulse 20 does not disturb any pixels read-out before or after the pulse, thus, the subtracted value is nearly zero. Accordingly, the pixel's value read-out during the flash heating pulse are elevated when they are subtracted from pixel values from non- heating pulse frames and the subtracted value will be greater than zero.
[0027] The subtracted image is effectively a real¬ time trace of the flash heating pulse intensity since the FPA is read-out row-by-row from top-to-bottom. For example, if a FPA camera has 200 rows and each row is read-out in 20 ms, and the higher pixel values are observed in rows 40-50, then the tfiash = (20 ms) (40+(50-4O) /2) / (200) =4.5 ms and the detectable flash duration is (20 ms) (50-40) /200=1 ms. Although subtraction (as mentioned above) is an effective method of comparing pre-flash pixel values with pixel values collected during the flash event, other method of mathematical manipulation (i.e. division, and the like) are also suitable alternatives.
[0028] A graphic manifestation of the implementation of the method of the present invention is set forth in Figures 3A, 3B, 4A and 4B. Specifically, the depiction of Figure 3A shows a graphic image constructed from data captured by a infrared camera using an approximately lms wide flash pulse. Using the techniques set forth above, it can be determined from this graphic representation that tfiaSh =0.97 ms and the duration 23 of the pulse event equals 1.04 ms. In contrast to Figure 3A, Figure 3B depicts a flash pulse event of approximately 1.04 ms in duration (same as Figure 3A); however, the onset of tfiaSh occurs at 11.02 ms
representing an onset delay of the flash pulse by approximately 10 ms later than that shown in Figure 3A. By comparing Figure 3B with 3A, tfiash can easily be determined using the methods disclosed herein without using anything other than information made available by the infrared camera. The graphic representation for extremely short durations will not manifest themselves as horizontal bands (as shown in Figures 3A-4B) but rather as one or more, adjacent illuminated pixels.
[0029] Figures 4A and 4B show a similar delay to that of Figures 3A and 3B (approximately 10 ms) , however, the flash duration in Figures 4A and 4B has been changed to 4.2 ms (as compared to the flash duration in Figures 3A and 3B which is approximately 1 ms) .
[0030] While various embodiments of the invention have been described herein in connection with the invention, it is to be understood that the embodiments disclosed herein are so disclosed by way of illustration and not by way of limitation. The scope of the appended claims should be construed as broadly as the prior art will permit.